The Australian government said it will put mandatory safety and security rules for frontier AI into law. Frontier AI means the most advanced models available. The rules would cover all frontier AI companies that want to operate in Australia. A new AI safety regulator would enforce them, working with the Australian AI Safety Institute. It could impose significant penalties. Companies would have to keep security controls in place from development to release. They would also have to manage serious risks, let outsiders evaluate their models, report incidents quickly, and publish their safety and security frameworks.
The sources do not give penalty amounts, a start date or the final text of the law. They also do not say whether results of independent evaluations would be made public. The government said more detail will follow once its work and consultation are done. It said fast incident reports should give earlier warning of risks from rogue AI agents.
The law would require a published framework and a model open to independent evaluation. Scope, penalties and timing are still open, and nothing says results would be shared. So buyers have no assurance from the law yet. They can ask vendors now for their framework and any evaluation evidence.